Robust Visual Tracking Based on Support Vector Machine and Weighted Sampling Method

Abstract Visual tracking algorithm based on binary classification has become the research hot issue. The tracking algorithm firstly constructs a binary classifier between object and background, then to determine the object’s location by the probability of the classifier. However, such binary classification may not fully handle the outliers, which may cause drifting. To improve the robustness of these tracking methods, a novel object tracking algorithm is proposed based on support vector machine (SVM) and weighted multi-sample sampling method. Our method constructs a classifier by sampling positive and negative samples and then to find the best candidate that has the largest response using SVM classifier. What’s more, the proposed method integrates weighted multi-instance sampling method, which can consider the sample importance by the different weights. The experimental results on many sequences show the robustness and accuracy of the improved method. The proposed target tracking algorithm in video target tracking with a variety of classic popular tracking algorithm, better able to achieve robust target tracking, but also in the infrared video, the infrared target tracking is also has the advantages of stable and accurate.